Soft Computing-Based Models for Estimating Undrained Bearing Capacity Factor of Open Caisson in Heterogeneous Clay.

Open caissons are commonly used in the construction of various underground structures, such as launch and reception shafts for tunnel-boring machines, storage or attenuation tanks, and cofferdams. During the sinking phase, the cutting edge of a caisson wall with a cutting face encounters soil and is subjected to loading to facilitate and control the sinking process. In this regard, the bearing capacity factor (N) of the cutting face of a circular open caisson in heterogeneous clay is evaluated via finite element limit analysis, which accounts for the increase in the undrained shear strength with depth. The parameters considered in this study cover practical aspects, including the excavation geometry, soil strength profile, and caisson geometry. This investigation also explored the impacts of the cutting face angle (β), roughness (α), ratio of the internal embedment depth to the embedment width (H/B), ratio of the internal radius to the embedment width (R/B), and strength gradient ratio (ρB/s_u0). In particular, when the H/B > R/B ratio, the N value tends to stabilize. Crucially, when H/B > 4, an increasing trend in R/B leads to a rise in N until R/B exceeds 10, i.e., large diameter caissons, stabilizing the N value. Furthermore, the results reveal the significant dependency of the cutting face roughness and strength gradient ratio of clay on N. The artificial neural network model, which is a soft computing-based model, is also developed to present the undrained bearing capacity forecasting equation. Compared with conventional regression, including the multiple linear regression model and the multiple nonlinear regression model, it has excellent performance, as measured by eight indices. In addition, ANOVA and Z-tests can support the research hypothesis and reject the null hypothesis. [ABSTRACT FROM AUTHOR]